Column

Home

Welcome to the Conservation Agents Leaderboard.

On this left side-panel, context is provided for the environments from the Fishing Gym. The right side-panel hosts the leaderboard where submitted agents are evaluated.

fishing-v0

This environment implements a simple, single species logistic growth-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow a logistic growth model.

Action Space The agent selects the amount of fish to harvest with respect to K. In this case, at interval of K/100ths: 0, K/100, K/50, 3K/100, K/25, …

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v1

This environment implements a simple, single species logistic growth-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow a logistic growth model.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v2

This environment implements a single species logistic growth-based fishery with a tipping point.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow a logistic growth model but below a population of 0.5K the population becomes much more likely to crash.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v4

This environment implements a single species logistic growth-based fishery with model error.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow a logistic growth model but with an r and K that are drawn from a normal distribution each episode.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v5

This environment implements a single species Allen model-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow the Allen growth model.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v6

This environment implements a single species Beverton-Holt model-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow the Beverton-Holt growth model.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v7

This environment implements a single species May model-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow the May growth model.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v8

This environment implements a single species Myers model-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow the Myers growth model.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v9

This environment implements a single species Ricker model-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow the Ricker growth model.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v10

This environment implements a single species non-stationary Beverton-Holt model-based fishery.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Dynamics follow a Beverton-Holt growth model where r changes constantly over an episode by some amount, alpha.

Action Space The agent selects a quantity of fish to harvest with respect to K with a continuous interval.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

fishing-v11

This environment implements a fishery where the transition dynamics vary each episode.

Observation Space The agent observes the fish population at that time step.

Model Dynamics Transition dynamics can follow May, Ricker, Allen, Beverton-Holt or Myers growth models. The dynamics model is randomly chosen every episode.

Action Space Dynamics follow the Ricker growth model.

Reward Function The agent is rewarded by the amount of fish harvested at a time step.

Column

fishing-v0

fishing-v1

fishing-v2

fishing-v4

fishing-v5

fishing-v6

fishing-v7

fishing-v8

fishing-v9

fishing-v10

fishing-v11